ui. - apps.dtic.mil
TRANSCRIPT
Ui.ASD-TR-87-5040
AD-A197 831
AN EVALUATION OF GROUND COLLISION AVOIDANCE SYSTEM ALGORITHM
Horace A. Orr, Major, USAFMr Richard GeiselhartMark DiPadua, Capt, USAFHuman Factors BranchGIMADS Program OfficeDirectorate of Support Systems Engineering
October 1987
rinal Report
Approved for public release; distribution unlimited.
DTICLECTE
AUG 2 2 18 8
H
DEPUTY FOR ENGINEERINGAERONAUTICAL SYST04S DIVISIONAIR FORCE SYSTEMS COMMANDWRIGHT-PATTERSON AIR FORCE BASE, OHIO 415433-6503
V
t,
NOTICE
IWhen Government drawings, specifications, or other data are used for any purpose otherth•n in conncction with a definitely related Government procurement operation, *.e UnitedStiates Governme.nt thereby incurs no responsibility nor any obligation whatsoever; and the factthat the government may have formulated, furnished, or in any way supplied the said drawings,specifications, or other data, is not to be regarded by implication or otherwise as in any mannerlicensing the holder or any othar person or corporation, or conveyinq any right or petrmssiori tomanufacture, use, or sell any patented invention that mpy in any way be related thereto.
This t•chnical report has been reviewed and is approved for piblication.
Progrm MaagerRalph j. Salvucc/Program Manager Chief, Crew Systems DivisionCrew Station Design Facility Directorate of Support Systems
Engineering
FOR THE COMMANDER:
If your address has changed, if you wish to be removed from our mailing list, or if theaddressee is no longer employed by your organization p!ease notify , W-P AFB,OH 45433 to help maintain a current mailing list.
Copies of this report should not be returned unless return is required by swurity considerations,contractual obligations, or notice on a specific document.
- UNCLASSIFIED
SECURITY CL.AMIFICATION OF THIS PAGEFr Approved
2b. DECLASSIFICATIONI/DOWNGRADING SCHEDULE is unlimited
4. PERFORMING ORGANIZATION REPORT NUMBER(S) 5. MONITORING ORGANIZATION REPORT NUMBER(S)
ASD-TR-87-5040
6a. NAME OF PERFORMING ORGANIZATION 6b. OFFICE SYMBOL 7a. NAME OF MONITORING ORGANIZATION(If applicable)
Crew Station Design Facility ASD/ENECH
6c. ADDRESS (City, State, and ZIP Code) 7b. ADDRESS (City, State, and ZIP Code)
Wright-Patterson AFB OH 45433-6503
Ba. NAME OF FUNDING'/SPONSORING 8b. OFFICE SYMBOL 9. PROCUREMENT INSTRUMENT IDENTIFICATION NUMBERORGANIZATION (If applicable)
8c. ADDRESS (City, State, and ZIP Code) 10. SOURCE OF FUNDING NUMBERSPROGRAM I PROJECT ITASK rWORK UNITELEMENT NO. NO. NO. ACCESSION NO.
11. TITLE (Include Security Classification)
An Evaluation of Ground Collision Avoidance System Algorithm
12. PERSONAL AUTHOR(S)Horace A. Orr, Maj, USAF: Richard GeiselharL; Mark A. DiPadua, Capt, USAF
13a. TYPE OF REPORT 13b. TIME COVERED ' 14. DATE OF REPORT (Year, Month, Day) 15. PAGE COUNT
Final FROM ,TO October 1987 78"16. SUPPLEMENTARY NOTATION
17. COSATI CODES 18. SUBJECT TERMS (Continue on reverse if necessary and identify bj block number)
FIELD GROUP SUB-GROUP Ground Collision Avoidance System Simulation Study
S17 07 . Terrain Estimation
19. ABSTRACT (Continue on reverse if necessary and identify by block number)
"'Controlled flight into terrain (CFIT) accidents, which constitute the second largestcategory of tactical Air Force Class A mishaps, could be greatly reduced by installingforward-looking Terrain-Following Radar (TFR) Systems. In an effort to design a lesscostly solution to this problem, a software-oriented generic Ground Collision AvoidanceSystem (GCAS) with application to more complex tactical missions was developed. Asimulation study of the system evaluated the adequacy of the algorithm in a variety oftactical scenarios. Results indicate GCAS is a viable system for partial solution ofCFIT. Although some modifications to the system are necessary pricr to implementation,the concept of terrain estimation may give GCAS a capability unmatched by any similar"systems.
20. DISTRIBUTION/AVAILABILITY OF ABSTRACT 21. ABSTRACT SECURITY CLASSIFICATION13UNCLASSIFIEDIUNLIMITED 0 SAME AS RPT. 0 DTIC USERS UNCLASSIFIED
22a. NAME OF RESPONI., :1 E INDIVIDUAL 22b. TELEPHONE (Include Area Code) 22c. OFFICE SYMBOLKatherine M. HoJnacki 513-255-5597 ASD/ENECH
DD Form 1473, JUN 86 Previous editions are obsolete. SECURITY CLASSIFICATION OF THIS PAGE
UNCLASSIFIED
TABLE OF CONTENTS
SECT ION PAGE
I INTRODUCTION 1
II STUDY PROCEDURE
1. Apparatus 3
2. Subjects 7
3. Procedure 7
4. Data Collection 9
5. Experimental Design 9
6. Missions 11
III STUDY RESULTS
1. Performance Data 26
2. Questionnaire Data 39
3. Algorithm Analysis 42
a. Altitude Loss-Pilot Response 43
b. Altitude Loss During Poll Recovery 47
c. Altitude Loss During Dive Recovery 49
d. Altitude Loss Due to Terrain Rise 53
4. Terrain Estimation 59
5. Recommer.dations 61
6. Results 62
IV CONCLUSIONS 63
APPENDIX
Pilot Questionnaire 64
Test Procedure 71
Mission Briefings 72
REFERFNCES 78
i iii
LIST OF ILLUSTRATIONS
FIGURE PAGE
1 Crew Station Design Facility 5
2 Functional Diagram of Crew Station Design Facility 6
3 Mission 11 14
4 Mission 11 (21) Low Level Navigation 17(Artist's Concept)
5 Mission 13 (23) Hard Turns at Low Altitude 18(Artist's Concept)
6 Mission 14 19
7 Mission 14 (24) Range Mission (Artist's Concept) 20
8 Mission 15 21
9 Mission 15 (25) Pop-Up Weapon's Delivery 24(Artist's Concept)
10 Mission 16 (26) IMC Maneuvering (Artist's Concept) 25
11 Frequency Distribution of Pilot Response Times 28for Control Input
12 Frequency Distribution of Warning Duration 32
13 Piecewise Calculation of Altitude Loss 42
14 Possible Nuisance Warning Caused by Failure to 47Project Flight Path Upward
15 G-Available vs Indicated Airspeed 50
16 Error Caused by Estimating ZtpuS Based on 52Recovery to = 0
17a Aircraft Receiving Nuisance Warning 56
iSbon For17b Aircraft Receiving No Nuisance Warning 56 CRA&I
TABI1a Linear Predictor Too Shallow for Downslope 57 Touneed Q
18b Linear Predictor Too Steep for Downslope 57
19 Late Warnings - Terrain Extrapolation 58tbution/
20 Nuisance Warnings - Terrain Extrapolation 58 tability Codes-Avail 'end/or
IDat 1Spo*Lal
iv
LIST OF TABLES
TABLE PAGE
I Pilot's Flying Time 8
2 Flight Parameters 10
3 Pilot Response Times (seconds) for Stick Input 27Following GCAS Warning
4 Analysis of Variance (ANOVA) of Pilot Response 27
Time (seconds) from GCAS Warning to Control Pilot
5 Mean Warning Duration Times 31
6 Analysis of Variance for Warning Duration Times 31
7 Mean Time to Maximum G's 33
8 Anaiysis of Variance for Mean Time to Maximum G's 33
9 Mean Maximum G's Across Missions 35
10 Analysis of Variance for Mean Maximum G's 35Across Missions
11 Mean Time to Zero Flight Path Angle 37
12 Analysis of Variance Mean Time to Zero 37Flight Path Angle
13 Summary of Crash Data 38
14 Questionnaire Data Summary 4]
v
SECTION I
INTRODUCTION
Although the Air Force accident rate in recent years has been at an
all time low, many feel that an even lover rate could be achieved by
reducing the number of controlled flight into terrain (CFIT) accidents,
which constitute the second largest category of tactical Air Force (TAF)
Class A mishaps. Air Force investigations concluded that between 1975
and 1981, there were 56 fighter and attack aircraft involved in CFIT
accident mishaps. The United States Navy has attributed the loss of 317
aircraft to CFIT accidents between 1970 and 1984. Of course, the occur-
rence of CTIT could be greatly reduced in the tactical fleet by in-
stalling forward-looking Terrain-Following Radar (TFR) systems, but such
a solution would be prohibitively expensive and possibly unnecessary.
Since a significant percentage of accidents occur during low level
missions while flying over sloping terrain that is not particularly
severe, it has been proposed that a less costly solution to the problem
may be a system based on less complicated sensors, such as a radar
altimeter system, that would cover a part of the CFIT envelope instead of
all of it.
Such an approach has been highly successful on comercial airliners
and some wide body military transports which are equipped with a Ground
Proximity Warning System (GPWS). The CFIT accidents for commercial
aircraft have in fact dropped to virtually zero since 1976, when GPWS use
was mandated by the Federal Aviation Administration. However, the much
more complex nature of tactical application cannot be accomodated by the
system currently available in commercial and wide-body aircraft. In an
effort to develop a Ground Collision Avoidance System (GCAS) which may
1
. i ~w .... .....................-...-- -
have application to the much more complex tactical mission, the Air Force
has contracted with Cubic Corporation to develop a generic GCAS system
that is soft-are oriented with a minimum of complex sensors required.
Such a system was developed and has been flight tested on a limited
basis. Due to safety considerations and in order to assess GCAS in a
wider variety of tactical situations, the CSDF was requested to conduct a
simulation study of the generic GCAS. The overall objective of the
simulation study was to evaluate GCAS in a variety of tactical scenarios.
More specifically, the purpose of the simulation was to provide the
following data to the GCAS program office:
a. Evaluate the adequacy of the algorithm to provide a warning
in sufficient time for a pilot to recover in a variety of
CFIT Scenarios.
b. Evaluate the impact and incidence of nuisance warnings gen-
erated by the algorithm.
c. Provide data to further refine and flight test the algorithm.
d. Assets pilot reaction to GCAS.
e. Provide normative data on pilot response time to GCAS
warnings.
2
SECTION 11
STUDY PROCEDURE
1. APPARATUS
The apparatus for conducting this experiment was the Crew Station
Design Facility, which has the capability to dynamically simulate a
complete flight regime under a variety of controlled conditions. The
facility consists of six basic components:
a. Three crew station shells - A-10, F-16, and C-135/C-18.
b. A digital computer complex consisting of five Gould Model
77/80SEL computers, one Gould Model 8780 SEL computer and A
Digital Equipment Corporation Model ?DP 11/34. All cockpits
are interfaced through a Singer Advanced Simulator TechnoLogy
(AST) interface.
c. A visual simulation complex consisting of a Singer Link NVS
Night Caligraphic Visual System and Singer Dual SMK-23 Citmera
Model Visual System.
d. Radar simulator equipment.
e. Monitors and Recording equipment.
f. Computer graphics complex.
The CSDF F-16 Flight Simulator was used as the test vehicle for
this study. An out-of-the-cockpit visual scene was provided using a
closed-circuit TV system from a modified Link SMK-23 moving terrain model.
The visual system consisted of a high resolution, low light level TV
camera and a Farrand optical probe, which transferred the mountainous
terrain images to a Conrac 1000 line, black and white TV monitor. The
scene was transmitted through a beam splitter and parabolic mirror with a
3
focal length of 54 inches. This provided apprximateoly a 48-degree-for-
ward field of view (yOV) to the pilot with the image collimated to appear
at infinity. The system provided simulated aircraft visual parameters of
360 degrees continuous heading, 360 degrees of continuous roll, plus or
minus 120 degrees of pitch, and 50 to 4000 feet altitude. All aircraft
abrodynamics and aircraft systeo4- were computed on the SEL computer com-
plex. The Voice Warning System wa- used to transmit digitized warnings
(pull-up, pull-up) to the pilot for each GCAS event. Objective data
parameters collected on each mission were taken from the SIM. computers
and recorded on magnetic tape for subsequent review and analysis.
Figures 1 and 2 provide an illustration and diagram of the apparatus.
The experimenter's console provides controls and displays for mission
setup, simulator operation, data collection equipment, and various
control and display mechanization options. For this evaluation there
were three opticns available to the experimenter to change simulator
position and attitude. One option allowed the console operator to move
the simulator forward 3000 feet in the flight plan to simulate a late
pop-up on weapon delivery runs. When this occurs in the real world, the
pilot has no difficulty finding his pull-up point, pulls up too close to
the target and ends up making his dive-bomb pass at a steeper than normal
delivery angle. This feature was used on missions 14 and 15 (auto pull-up
missions 24 and 25) to complicate the delivery task.
Another option allowed him to decrease simulator altitude in
increments of 1000 feet. This was used most often on mission 16 (auto
26) to simulate entry into dynamic maneuvers at a lower than expected
altitude. This led to unexpected GCAS warnings in attitude and flight
path conditions approaching the limits of system operation and in some
cases, disorientation.
4
__________ -- - - - - - -�w--w--- '------- .-- - - ---- - -
4�'I
NbS.
'V
-I"-a
".4
oJ
-I
l.a
".4
5
IL
£V3
So..
N4)
.J
CJ
6U
The final option dealing with simulator control was the mechani-
zation of a joy stick on the console to operate in series with the
pilot's control stick. With this capability, the experimenter could
change aircraft pitch and roll attitudes during periods of pilot distrac-
tion to simulate entry into unusual (or unexpected) attitudes. This
control feature was used effectively during low level navigation anid high
"G" turning flight when pilots were instructed to read authentication and
alphanumeric cards.
2. SUBJECTS
A total of 10 subject pilots participated in the study. Eight of
the pilots were operational pilots from the Tactical Air Command (TAC),
and two were pilots from Wright-Patterson AFB with previous tactical
experience in the low level environment. The pilots had varying amounts
of experience in the F-16 as indicated in Table 1. They arrived over a
2-1/2 week period with an average of four pilots participating each
week (two subjects over a 2-1/2 day period).
3. PROCEDURE
When the pilots arrived to participate in the study, they filled out
a personal data questionnaire (see Appendix) and were given a general
overall briefing on the program. A detailed briefing on the GCAS algo-
rithm was also presented along with "ground school" on the F-16 sim-
ulator. The pilots flew a training mission in addition to the briefings
to conclude the inital half-day period. In the er'uing day and a half,
the pilots flew 12 missions in which various data described earlier were
collected. In the last half-day block, each pilot was debriefed and
filled out a questionnaire. A detailed schedule of events for each pair
of pilots is in the Appendix.
7
TABLE I
TAIIt- 1
PILOT'S FLYING TimsSUBaCT TOTAL TOTALNUMUR FLYING TINE (us) 7-16 TLIN (nS)
1 350 150
2 400 110
3 1500 1250
4 1700 350
5 600 350
6 300 200
7 1780 780
8 1700 640
9 2000 600
10 1170 0
M- 1155 443
Pilot briefings prior to each data collection flight covered the
overall mission scenario, specific objectives of the mission, altitudes,
speeds, times, etc. Pilots were also provided with a mission setup
checklist and a strip map of the route of flight when appropriate.
Mission briefing outlines are provided in the Appendix. Following the
briefing, each pilot was taken to the simulator to fly the test mission.
When the pilot had completed cockpit setup actions, previewed
the mission and indicated that he was ready to start the run, the
simulator and data recording equipment were activated and the mission
scarted. Voice communicationa were held to a minimum during each
evaluation run and limited to responding to the pilots' questions, or as
8
was the case with missions 11 and 15 (auto 21 and 25), recording his
assessment of the validity of GCAS warnings, since these were the only
missions where he was in a position to assess validity of the warning as
valid, invalid (expected), or invalid (unexpected).
Following each simulated mission, a short debriefing session
provided an opportunity for the experimenter and evaluation pilot to
discuss simulator and GCAS operation. During these sessions, the experi-
menter recorded pilot comments on GCAS operation as it applied to the
mission just flown. The questionnaire with pilot comments can be found
in the Appendix.
4. DATA COLLECTION
Table 2 shows the flight parameters and performance measures that
were collected during all test missions. These parameters were to recon-
struct the flights for more definitive evaluation. The critical flight
parameters were plotted as time histories for each maneuver causing a
warning. The data were plotted from a period of 5 seconds prior to
the warning to the minimum recovery altitude above ground. These data
plots were used to riconstruct each GCAS event for data analysis and
assessment of pilot performance.
5. EXPERIMENTAL DESIGN
The experimental design was a treatment x treatment x subjects
(2x2jclO) or ABS design as described in Lindquist's D~esign and Analysis of
Experiments in Psychology and Education (Ref 1). The treatment
variables were (1) type of recovery (manual or automatic) and (2) type of
mission (six representative tactical missions). The manual pull-up data
were used to assess pilot performance and the automatic pull-up data were
to analyze the algorithm (i.e., provide consistent response without
9
TABLE 2 FLIGHT PARAMETERS
FLIGHT PARAMETERS UNIT OF MEASUREMENT RANGE
Pitch Attitude degrees +90
Roll Attitude degrees +180"G" load "$GO -3 to +10
Flight Path Angle degrees +90
Pitch Rate degrees/sec +45
Absolute Altitude feet 0-1000
Barometric Altitude feet 0-6000
Heading degrees 0-360
Roll Rate degrees/sec +75
Yaw degrees +45
Terrain Angle degrees +60
GCAS triggered Pull-up time sec Discrete
Altitude at GCAS alarA feet Discrete
Altitude clearance at feet Discretepull-up instance
Engine RPM %RPM Idle-95
Groundspeed knots 0-600
True Airspeed knots 0-600
Angle of Attack degrees
Stick Force lbs Discrete
Time from alarn to change in clock Secondsstick force (response time)
10
variations in response time).
6. MISSIONS
Six missions were designed to test the algorithm. Overall
objectives in designing the profiles were to subject the pilot to a
variety of conditions representative of those encountered in tactical
fighter operations with specific emphasis on the kinds of situations
reported in CFIT incidents. A complete description of each mission
follows:
1. Low Level Navigation - Mission 11: This micsion was flown at
300 to 500 feet above the terrain. Some segments of the route were
level terrain, between ridges, simulating typical terrain masking oper-
ations while other segments required ridge crossings and flight over
rough terrain (Figures 3 and 4).
2. Low Speed Maneuvering IMC - Mission 12: This mission was designed
to simulate radar vectors for approach or low level, low specd,
navigation where the controller or the pilot selected a maneuvering
altitude below minimum safe altitude and slightly below the level of
ridges along the route of flight. Flown at 1500 feet MSL (altitude below
ridgetops), 250KCAS and in instrument meteorlogic.l conditions, this
mission looked at the algorithm in a nearly level flight, wings level
condition where the pilot had no external visual cues from outside
terrain.
3. Hard Turns at Low Altitude - Mission 13: This mission was flown
over level terrain at 480KCAS in a figure-eight pattern between two
steerpoints. At each steerpoint passage, the pilot initiated a 5 'C' turn
of about 210 degrees to simulate defensive maneuvering at low level
300 to 500 feet). During the turns, pilots were asked to read a series
11
of numbers and letters from a card located on the canopy over his head to
simulate checking 6 o'clock for other aircraft. Adjustments to aircraft
attitude were made by the experimenter during the turns on this mission to
further complicate the recovery task (Figure 5).
4. Range Mission - Mission 14: This mission consisted of a series
of three runs on the same target from different approach heading. simu-
lating a low level range familiarization flight. The profiles were flown
at 300 to 500 feet and 480 KCAS on the low level navigation segments.
Delivery maneuvers consisted of a 30-degree straight ahead pop at 20 sec-
onds prior to the target, a climb to 4000 feet MSL and a roll to inverted
flight for the pull down to a 30-degree dive. Bomb release was briefed to
be completed at 2500 feet MSL. To complicate the delivery, the experi-
menter could move the aircraft (simulation) forward 3000 feet during the
a steeper then desired, or expected, dive angle that had to be used to
bring the target into view (Figures 6 and 7).
5. Low Level Navigation/Pop-Up Deliveries - Mission 15: The low
level portions of this mission were essentially the same as described in
* Mission 11 above except that three targets were selected along the route
of flight for pop-up weapons delivery maneuvers. Like the range
mission above, the pop was initiated 20 seconds prior to the target for a
30 degree delivery. Again, the experimenter could advance the siimulator
in the pop so that the aircraft would be outside delivery parameters in
the pull-down (Figures 8 and 9).
6. IMC Maneuvering - Mission 16: This mission flown in INC was made
up of a series of aerobatic maneuvers. These maneuvers required the
pilot to recover from relatively extreme pitch and bank attitudes such as
might be encountered in unusual attitude or spatial disorientation inci-
dents (Figure 10).
12
On some of the manuevers such as the loop, split a, and Cuban 8, the
experimenter adjusted siumlator altitude 3000-4000 feet to cause early
GCAS alarm.
13
II
14
I L I
~'igure 3a. Mission 11 (continued)15
~ IJ * A
f *
Niuel. iso 1 x
I
t4J0.IIU
0C.)
.9.'4)'.4
0
4.)
Co'9.4
CiKCi
.-1
30
-S
-. 4
N4 N 5-
0r4
Co.9-4
z
-t
41�.4
CO.9-4�L1
(
, (I
(I pI " iI
..
I IK.i
C"ii
Figure~~ 6risin1
19 .l
/
I./
/
I4JU"4'a
'5-.
0
U)
*1*4U)Pa
H 00
A"5--5'
4
- 5-
.4.
0.1.4U)U)
*1*4
v-s
0)I..
00*1.4
K
00-I
Figure 8. mission 15
21
I if~
Figure 8a. Mission 15 (continued)
22
0;3
Figure 8b. Mission 15 (continued)
23
PI
r-1-
0
0
1-1
ýlj 24
41$4.
00
r.4
1-1
0
00
.P4
250
SECTION III
STUDY RESULTS
1. PERFORMANCE DATA
Table 3 shows the mean pilot response time for making a control
input following a GCAS warning for individual missions and across all
experimental conditions. Of 220 warnings there were 208 warnings that
elicited a control response from the pilots. The 12 warnings that did
not result in a control input were too brief for pilot response or were
ignored for various reasons that will be discussed later. The overall
mean response time was 0.66 seconds with a standard deviation of 0.30.
The size of the standard deviation indicates a departure from a
normal distribution. In order to determine the extent of the departure,
a graph of the frequency distribution was drawn (Figure 11), and it shows
a positive skewness in the distribution (shaded area) of pilot response
times. Further inspection of these data reveals that approximately 80
percent of the area (10 out of 13 cases) in the tail of the distribution
is from mission 16 (INC/unusual attitudes), where the pilots were induced
into a disorienting environment, which no doubt contributed to the longer
and somewhat atypical response times. Almost 25 percent of the experi-
mental ruas on mission 16 resulted in pilot disorientation which was more
frequent than initially thought possible and provided valuable data for
an&lysis of GCAS. In order to compensate for the skewed distribution, the
median and semi-interquartile range (Q) were computed to use as a basis
for determining the time allowance for pilot response time in the GCAS
algorithm. In actuality, the difference between using the mean and
median in this case is probably only of academic interest. If you use
26
TABLE 3
TABLE SHOWING PILOT RESPONSE TIMES (SECONDS) FOR STICK INPUT
FOLLOWING GCAS WARNING
MISSION MEAN TIMA (SECONDS) S.D. N
11 (LO LEVEL NAV) 0.53 0.31 30
12 (IMC/LOW SPEED) 0.62 0.28 30
13 (LO ALT TURN) 0.57 0.22 22
14 (I-LNGE) 0.66 0.26 54
15 (LO LEVSL 0.64 0.27 26BOMB/NAV)
16 (tHC UNUSUAL 0.86 0.39 46A'.TITUDES)
ACROSS ALL MISSIONS: H - 0.66 S.D. 0.30 N i 208
TABLE 4
ANALYSIS OF VARIANCE (ANOVA) OF PILOT RESPONSE TIME (SECONDS)
FROM GCAS WARNING TO CONTROL PILOT
SOURCE SUM SQ DEGREES OF FREEDOM MEAN SQ F-RATIO
MISSION NO. 24.6 5 4.92 5.38
ERROR 184.8 202 0.91
TOTAL 209.4 207
*LEVEL OF SIGNIFICANCE < 0.001
27
"4$
"440
C,4
ue '44
* 0
02
co
* 1-4
0
en v~ "4 U, (4 0% %0 eNI~ "4N V4 "-4 "4
sesU ;o -rxpzwrI
28
the median to generate a pilot population value, the recommended time to
allow for pilot response time would be 1.11 seconds (0.59 median value
plus twice the 0.2o Q value) versus 1.25 using the mean and standard
deviation (0.66 mean value plus twice the 0.30 S.D. value). The authors
reco mme nd the 1.25-second interval in this case to allow for the
possibility of pilot disorientation or distraction in some situations.
This probably represents a good compromise between allowing the operator
some flexibility without a significant increase in nuisance warnings.
In comparing the mean response times in Table 3, we see that
missions 11 and 13 yield the lowest average response times (0.53 seconds
and 0.57 seconds respectively), followed by missions 12, 14, and 15 with
slightly higher response times (0.62, 0.64 and 0.66 seconds) and finally
mission 16 with a relatively large mean response time" (0.86 seconds).
Analysis of variance performed on theme data indicate the differences are
statistically significant at less than the 0.001 level (Table 4). The
significance was undoubtedly due to mission 16, which was discussed
earlier.
The standard deviations for the individual missions in Table 3 are
slightly larger than you would get in a normal distribution showing
increased variability and possibly some skewdness. The distributions for
mission 11 through 15 are symmetrical, with milssionl 16 showing the skew
previously mentioned.
Of the twelve warnings that did not elicit a response, two were too
brief for-the pilot to respond, four were ignored because the pilot was
in the midst of weapon delivery, two were coincidental with the pilot's
recovery, and four warnings were considered invalid.
A summary of warning duration times is shown in Table 5 and provides
some interesting insights on GCAS. Mean warning duration was 3.1 seconds
29
with a standard deviation of 2.34. The relative size of the S.D. is even
greater than in the response tuie data, indicating even more variability
and skewdness (Figure 12). In this instance, the skew is generated from
two sources - mission 12 and mission 16. The reason for the larger mean
and S.D. on mission 16 (induced diaorientation) was discussed earlier and
requires no further comment.
Mission .-2 was a low speed maneuvering flight under instrument
meteorlogical conditions (IMC) and was inserted into the study to assess
the low speed characteristic of GCAS. The extremely large mean 5.60-
second warning duration is due to the fact that all warning situations on
mission 12 resulted in a crash and the conditions necessary to silence
the voice warnings were never fulfilled. The crashes are attributed to
the method of computing terrain rise, altitude loss and the resulting
pilot response times. A comprehensive analysis of the algorithm is
covered in greater detail later in the results section. It is sufficient
to say at this point that changes to the algorithm will be required to
afford some protection in the straight and level flight regime. One
final point on mission 12, an offline analysis of the algorithm showed
that slope of the terrain was not a factor in any of the crashes. The
Analysis of Variance of the warning duration data (Table 6) verifies the
mean differences in tables are significant and are obviously due to
missions 12 and 16.
One of the factors that can be crutial to the s'ccess of a GCAS is
how long it takes a pilot to get maxiaum G's on the aircraft - the
quicker you can get to maximam G, the less altitude lost. Table 7
summarizes the mean time to maxirmum G data. The pattern of these data is
similar to measures discussed earlier - slightly larger than average
variability, and some skewdness with mission 16 accounting for most of
30
TABLE 5
MEAN WAARKIN DURATION TIMES
MISSION MOAN TIME(SCONOS) S.D..
11 (LO LEVEL) 1.85 1.08 32
12 (IMC/LOW SPUD) 5.68 1.39 31
13 (LO ALT TURN) 1.50 0.94 22
14 (RANGE) 2.32 2.19 58
15 (LO LEVEL 2.28 1.47 26BoMB/NAV)
16 (r14C UNUSUAL 4.38 2.49 51ATTITUDES)
ACROSS ALL M'ISSIONS, M - 3.10 S.D. - 2;34 N - 220
TABLE 6
ANALYSLS OF VARIANCE FOR W4ARNING DURATrON TIMES
SOURCE SUM SQ DEGREES OF FREEDOM KEAN4Sq F-RATIO
M.SSION NO. 4986 5 897 25.59
ERROR 7503 214 35
TOTAL 11989 219
* LEVEL OF SIGNIFICANCE < 0.001
31
0
00
00
CD Ch
I- q-
LL
~,won
IA IC
ult N
sasi3o ;o iaq
32L
TABLE 7
HEAR TIME TO MAXIMUM ,'S
MIsSION MEAN TIME(SECONDS) S.D. N
11 (LO LZEVL NAv) 1.86 0.89 32
12 (IMC/wW SPUED) 2.73 1.42 31
13 (Lo ALT TURN) 2.16 0.63 18
14 (RANGE) 2.27 1.14 56
15 (LO LEVEL 2.62 1.11 25BOMB/NAV)
16 (IMC UNUSUAL 3.80 1.50 51ATTITUDES)
ACROSS ALL MISSIONS: M - 2.68 S.D.- 1.39 N - 213
TABLE 8
ANALYSIS OF VARIANCE FOR MEAN TIME TO MAXIMUM G'S
SOURCE SUM SR DEGREES OF FREEDOM M F-RATIO
MISSION NO. 991 5 198 13.20
ERROR 3093 207 15
TOTAL 4084 212
* LEVEL OF SIGNIFICANCE < 0.001
33
it. The overall mean time to maximum G is 2.68 seconds with a S.D. of
1.39. ?Zissions 11 through 15 are grouped with small differences between
them, while mission 16 is significantly larger (Table 8). We should be
somewhat cautious in interpretting 'G' onset data in fighter simulators,
since they cannot simulate the phymical sensation of anything above I G.
It is possible that 'G' onset may be quicker in a simulator than in an
actual aircraft. However, the F-16 force stick would, in the opinion of
the authors, more likely be representative than a simulator with a
displacement stick. In any case, the simulator data is probably
representative of the actual aircraft, but flight test data should be
used to verify this.
Table 9 depict. mean maximum C across missions. The mean values
range from 2.32 for mission 12, to 5.9 for mission 16. Unlike other data
distributions, maximum G standard deviations are quite small in relation
to their respective means, reflecting normal sy-mmetrical and tight
groupings. Mean differences between all these individual missions are
significant (Table 10). The maximum G data reflects the maneuvers
required on each mission. The relatively low G's on mission 12 (2.32)
are all the G's available to the pilot on a low speed zero flight path
angle mission; whereas 5.9 G mean seen in mission 16 is representative
of IMC acrobatic mission. Although we must not take simulator G data too
literally, the figures in this case appear to match those typically
encountered in actual aircraft.
The final group of flight parameter data analyzed was mean time to
sero flight path angle (FPA) (Table 11). The total N in this case was
somewhat reduced since the data from mission 12 were not applicable,
since zero FPA was never achieved. The only significant data showM are
that the mission 16 mean is mauch larger than the others, b-.h because
34
TABLE 9
MEAN MAXIMUM G'S ACROSS MISSIONS
MISSION MAXIMUM G S.D. N
11 (LO LEVEL NAY) 3.38 0.98 32
12 (IMC/LOW SPEED) 2.32 0.34 31
13 (LO ALT TURN) 5.52 1,62 22
14 (RANGE) 4.86 1.81 58
15 (LO LEVEL 5.15 1.96 26BO tE / IAV)
16 ([11C AJSUAL 5.90 1.11 51ATTITUDES)
ACROSS ALL MISSIONS: M 4.62 S.D. - 2.34 SN" 22
TABLE 10
ANALYSIS OF VARIANCE FOR MEAN MAXIMUM G'S ACROSS MISSIONS
SOURCE SUM SQ DEGREES OF FREEDOM MEAN Sq F-RATIO
MISSION NO. 3270 5 654.0 32.22
ERROR 4350 214 20.3
TOTAL 7620 219
* LEVEL OF SIGNIFICANCE < 0.001
35
of the complex maneuvers and the induced disorientation (Table 12).
Almost 30 percent of the GCAS events observed in the study rcasilted in
simulated crashes. A statistical summary of these crashes is shown in
Table 13. Almost one-half the craches occurred on mission 12 and were
due to the way the GCAS algorithm is mechanized. This will be e"overed in
the Algorithm Analysis. Of the remaining 33 crashes, 21 occurred on
bombing type missions where the pilots tended to ignore the warning until
weapons delivery was completed, and six occurred during aerobatics where
disorientation was a factor.
One reason pilots flew through the warnings involved how frequently
they occurred at an obviously high altitude, especially at the higher
dive angles where pilots were trying to release at the briefed altitude
of 2500 feet. When warnings occurred at 3000 feet and abov~e, they would
have to make a judgement: was the warning early or was it caused by a
higher than normal dive angle? This required time and concentration at a
point in the delive~ry where intense concentration was required as the
delivery solution. This problem is not unique to the air-to-ground
delivery mission; in fact, it is one faced by pilots on almost every run
in an unfamiliar target area. In an operational sense, the tendency to
fly through a warning will surely be less in an aircraft than it was in a
simulator, and pilots will at least take a second look at the situation
when the warnings occur. In this same view, however, the system will
lead to occasional early pull-ups and sub.sequent dry passes.
Probably the most significant problems were those encountered in
mission 12, where all warnings led to crashes and on other missions where
rising terrain was encountered in a near level flight attitude. In these
cases, pilots received late warnings or none at all. The underlying
causes of this problem are described in Section 3, Algorithm Analysis
36
TABLE 11
MEAN TIME TO ZERO FLIGHT PATH ANGLE
MISSION MEAN TIME (SECONDS) S.D. N
11 (LO LEVEL NAV) 2.06 1.19 22
12 (IMC/LOW SPEED) N/A N/A N/A
13 (LO ALT TURN) 2.36 0.79 22
14 (RANGE) 2.72 1.30 38
15 (LO LEVEL 2.73 1.73 21BOMB/NAV)
16 (IMC UNUSUAL 7.20 1.73 43ATTITUDES)
ACROSS ALL MISSIONS: M - 3.88 S.D. - 2.56 N - 146
TABLE 12
ANALYSIS OF VARIANCE MEAN TIME TO ZERO FLIGHT PATH ANGLE
SOURCE SUMSQ DEGREES OF FREEDOM MEAN Sg F-RATIO
MISSION NO. 67650 4 16912 89.7
ERROR 26583 141 188
TOTAL 94233 145
* LEVEL OF SIGNIFICANCE < 0.001
37
TABLE 13
SU001ARY OF CRASH DATA
PERCENT PORCENTNO. CRASHES NO. WARNINGS OF MISSION OF TOTAL
MISSION 11 5 32 16 8
MISSION 12 31 31 100 48
MISSION 13 1 22 5 2
MISSION 14 14 58 24 22
!11SS310N 15 7 26 27 it
MISSION 16 6 51 12 9
TOTAL: •4 220 NA 100
38
TABLE 11
MEAN TIME TO ZERO FLIGHT PATH ANGLE
MISSION MEAN TIME (SECONDS) S.D. N
11 (LO LEVEL NAV) .2.06 1.19 22
12 (IMC/LOW SPEED) N/A N/A N/A
13 (LO ALT TURN) 2.36 0.79 22
14 (RANGE) 2.72, 1.30 38
15 (LO LEVEL 2.73 1.73 21BOMB/NAV)
16 (IMC UNUSUAL 7.20 1.73 43ATTITUDES)
ACROSS ALL MISSIONS: M - 3.88 S.D. = 2.56 N = 146
TABLE 12
ANALYSIS OF VARIANCE MEAN TIME TO ZERO FLIGHT PATH ANGLE
SOURCE SUM SQ DEGREES OF FREEDOM MEAN SQ F-RATIO
MISSION NO. 67650 4 16912 89.7
ERROR 26583 141 188
TOTAL 94233 145
* LEVEL OF SIGNIFICANCE < 0.001
37
TABLE 13
S001ARY OF CRASH DATA
PERCENT PERCENT
NO. CRASHES NO. WARNINGS OF MISSION OF TOTAL
MISSION 11 5 32 16 8
MISSION 12 31 31 100 48
MISSION 13 1 22 5 2
MISSION 14 14 58 24 22
MITSSION 15 7 26 27 ti
M(SSION 16 6 51 12 9
TOTAL: 54 220 V/A 100
38
(in 3.a Condition 1 and 3.c). Basically, the problem Sterns from system
inhibits installed to prevent nuisance warnings during ridge crossings
and it is one which will be extremely difficult to solve. If the
inhibitte are removed to provide a "look ahead" or terrain projection
capability, nuisance warnings go down; if the inhibits are used, there
is little or no protection against rising terrain.
2. QUESTIONNAIRE DATA
Pilot ratings on all aspects of operacional utility were
overwhelmingly favorable towards GCAS. These ratings are summarized
in Table 14. Eight out of ten pilots thought the warnings were timely
and the system was either useful or extremely useful in terms of system
utility. These same pilots also said they got nuisance warnings
occasionally, did not tend to over rely on the warnings, and missed no
warnings. All participants agreed the attention getting value of GCAS
was good. Although all pilots thought the system was either good or
satisfactory, they all agreed some improvements were required.
In the way of system improvement, there was general agreement (6
out of 10) that ridge crossings presented a problem and require an
algorithm to modify and improve the GCAS in this situation. Two pilots
reported that they would like a system with some foward looking
capability. Some sort of anticipatory cue prior to the warning was
suggested by three of the pilots. Finally, one pilot suggested a "more
timely" voice.
General comments made by the pilots were all favorable and two
thought GCAS should be installed immsediately. Six out of ten pilots
thought the slope warning unnecessary, two thought it moderately useful,
39
71
one preferred it, and one made no coimment. Finally, the pilots generally
agreed that the sjimulation warn satisfactory for rating GCAS. If the
reader is interested in reading all pilot comments, they are provided in
the Appendix.
40
TABLE 14
QUESTIONNAIRE DATA SUMMARY
TIMELINESS:
2/10 SLIGHTLY EARLY7/10 JUST RIGHT1/10 SLIGHTLY LATE
ATTENTION GETTING VALUE:
5/10 ALL THE TIME5/10 MOST OF THE TIME
NUISANCE WARNING OCCURRENCE:
8/10 OCCASIONALLY2/10 MOST OF THE TIME
MISSED WARNINGS:
8/10 NONE2/10 YES
SYSTEM UTILITY:
8/10 USEFUL OR EXTREMELY USEFUL2/10 KOOERATE USEFULNESS
OVER RELIANCE ON WARNING:
8/10 SELDOM OR NEVER1/10 OCCASIONALLY1/10 FREQUENTLY
ADEQUACY/SUITABILITY:
8/10 SATISFACTORY, Bur NEEDS IMPROVEMENT2/10 GOOD OR VERY GOOD
41
3. ALGORITHM ANALYSIS
The GCAS algorithm was analyzed by CSDF in order to determine the
validity of warnings in various flight conditions. For the purpose of
this evaluation, an invalid warning is one which occurs too early (a
"1nuisance" warning), or one which occurs too late for a safe recovery.
The algorithm is briefly explained here in order to provide a background
for the analysis. Cubic's algorithm continuously compares the aircraft's
present height above the ground with the altitude required for dive
recovery. The algorithm predicts altitude loss during dive recovery as a
sum of the following:
a. Altitude loss due to terrain rise during the pull-up.
b. Altitude loss due to pilot response time.
c. Altitude loss during roll recovery.
d. Altitude loss during a 5-C pull-up.
Figure 13 illustrates this piecewise calculation of altitude loss. - -
I TERRAIN d... -. . .EXTRAPOLATED -- - ....
FORWARD IN- ---TiME
TOTAL ALTITUDE REQUIRED= I+II+III+IV+VIII ALT. LOST DUE TO H
RESPONSE TIME T1
III ALT. LOST DUE 10ROLL RECOVERY
IV ALT. LOST DUE TODIVE RECOVERY
V TERRAIN CLEARANCE LIMIT
Figure 13. Piecewise Calculation of Altitude Loss
42
The following paragraphs describe each portion of the altitude loss
calculation in detail. Reasons f or early or late warnings are also
dis cussed.
a. Altitude loss-pilot responbe
The initial portion of the altitude loss is fairly simple to
compute. During the period of time before the pilot has made any
response to a warning, the aircraft is projected to continue its present
vertical trajectory. The current vertical velocity is multiplied by a
computed response time (T rep) to calculate altitude that would be lost
from warning to initial stick input.
NOTE: If the altitude loss is negative (i.e., the airplane climbing the
absolute value of the altitude loss is used), Cubic assumed a basic
response time of two seconds and modified it by several conditions:
Condition 1 : Small Dive Climb Angle (GAMMA)
In order to reduce the possibility of early warnings during ridge
crossings T repis reduced when the aircraft is in level or near level
flight. The following relation is used:
If < 5'
T 0n.4reap
So at zero flight path angle, T ra- 0
Condition 2 : Roll Angle/Roll Rate (&0)is assumed that if the
pilot is rolling out of a bank at the time a warr~ing occurs, his reaction
time will be less, so:
If 0> 15 deg/sec
and
J6 and are of opposite sign,
T reap 0.3 sec
43
NOTS: This condition overrides Condition 1.
Condition 3 : Vertical Velocity, Vertical Acceleration. Response
time is assumed to be shorter when the pilot is already reducing his
downward velocity. So, whenever the aircraft has a downward flight path
(Va <0) and a positive acceleration vertical acceleration (A <0), the
response time used in calculating the airplane's future flight path angle is
reduced by half. Otherwise, if the airplane has a downward acceleration,
its vertical velocity will be modified using full response time as calculated
by Conditions I and 2. So the extrapolated dive/climb angle, X ext is
calculated as follows:
'Y'ext 0 tannl V (ex)
where Vs(ext) = V + A (T ext extrapolated vertical velocity)
Text = Treap (modified by Condition 3)
Vx - Horizontal Velocity
The value Xext is used in computing the altitude required for a 5-G
(or available G) dive recovery (Paragraph C). If the airplane has a
positive vertical velocity and a positive vertical acceleration, the
extrapolation is not performed. Additionally, if Vext > 0, no
(positive) altitude change will be calculated due to pilot response
time.
Condition 4 : Previous warning within 3 seconds. It is qssumed that
if a pilot has recently (within the past 3 seconds) experienced a pull-up
warning, he will react to subsequent warnings more quickly. Thus, if the
warning is not on and less than 3 seconds have elapsed since the last
warning ceased,
T - 1.5 secondsreap
Condition 5 : Warning in progress. When a pull-up warning is in
44
effect, the response time is reduced by one time stop each time the
calculation is per foruad.
So, for the basic 2 second response time, T repvill be reduced to zero
after 2 seconds of warning.
a.l1 Problems
a.1.l 1Karly warnings
a.1l.1. The basic response time of 2 seconds is too long. A
more realistic baseline would be between 1 and 1.25 seconds (see
performance section of results).
a.1. 1.2 The establishment of a 0.3 second response time under
Condition 2 is intended to reduce nuisance warnings. In some cases,
however, it actually causes an increase in invalid warnings. Consider
first the case where a pilot gets a warning during descending turn
( V W5). His response time (T rep) starts at a value of 2 seconds.
As he begins to roll out of his turn, Condition 2 immediately reduces
his response time to 0.3 seLonds. This will often cause the alarm to
terminate since altitude loss due to response time has decreased.
However, as soon as (a) his roll rate falls below 15 degrees per second
or (b) his bank angle changes sign because he overshoots wings level,
the response time resets to 1.5 seconds (Condition 4). This can cause
the warning to come back on even though he has begun a satisfactory
recovery. Modifications to altitude loss based on roll rate would perhaps
be more appropriately applied to the roll recovery portion of the
algorithm.
It would also be prudent to suppress response time reset (Condition 4)
completely for 3.0 seconds after a warning has been satisfied. As a
minilmum, it should be reset to a much smaller value than 1.5 seconds.
45
a.1.l.3 It is not valid to assume, as in condition 1, that a
pilot's response time is less in level (%~ a 0) flight. The response
times demonstrated in this study shoved no substantial reduction under
these conditions. In level flight, the only source of a ground
collision can be ground rise. Paragraph a. shovs that this
algorithm does not consider pilot response time in calculating altitude
loss due to terrain rise. Therefore, Condition 1 provides only a
negligible reduction in nuisance warnings during ridge crossings.
a.1.1.4 Condition 3 modifies the airplane's predicted
altitude loss according to instantaneous vertical velocity and
acceleration. It can cause abrupt changes in the predicted altitude
loss as the aircraft transitions from one set of conditions to another,
resulting in erroneous warning resets in a manner similar to the
example in paragraph al.l.2. Also, since the flight path is never
extrapolated upward during a climbing maneuver, some needless warnings
may occur (Figure 14).
Consider the case where a pilot is pulling up to avoid a
ridgeline. The algorithm projects zero altitude gain during the response
time. So even though the pilot may be climbing and accelerating upward,
he will get a warning based on the fact that his current altitude is leass
than the projected terrain height. This is compounded by the fact that the
time used for calculating terrain rise is based on dive recovery time. A
positive dive recovery time will be calculated even though the airplane's
flight path is above the projected terrain (Paragraph d).
46
-O-- .
Linear (1G) predicted fLight path
Acttual (3G)predicted fLight Path
3~~~ ~ GE ---- Light path predicted-7 -by aLgorithm
Figure 14. Possible Nuisance Warning Caused by
Failure to Project Flight Path Upward
A more accurate way to approximate the altitude loss during pilot
response time would be to simply apply the following equation to obtain
altitude lost during pilot response. That is:
A zreap "-V Cos ( )
where:
m Present flight path angle
S2 u Y 1 +. ý 1 (T reap)
V - Aircraft Velocity
and, assuming relatively small change in ,1I g (nv -cosY )
Vnv n cos
This simple equation provides both altitude loss (or gain) during pilot
reaction time and a predicted flight path angle to be used in the dive
recovery equations. It eliminates abrupt changes in predicted loss and
resultant nuisance warnitigs that are characteristic of the current
algorithm.
b. Altitude Loss During Roll Recovery
The GCAS algorithm predicts the altitude that will be lost while
the pilot rolls wings level. This calculation is based on the assumption
that the pilot will not begin his dive recovery until he has brought his
47
wings nearly level. The calculation is made in the following manner:
(1) The time for roll recovery is calculated.
T rr 20
rpred
Where * roll angle
Spred a predicted roll rate
- 70 /Sec for 20 ° < 0 < 85
56 "/Sec for 85 " < 0 <145
90 °/Sec for 145 *< 0 <180
(2) The vertical velocity is extrapolated
V -V +A *T
rr z z rr
where
V - Extrapolated vertical velocityrr
V A Instantaneous (measured) vertical velocityz
A - Instantaneous (measured) verticalZ acceleration
(3) The flight path angle is extrapolated:
Y rr tan- Vrr
Where V x Instantaneous horizontal velocityx
(4) The altitude loss due to roll recovery is calculated.
rr z rr
If the altitude loss is negative (i.e. a climb), the absolute
valuelA Zrr is used.
If the extrapolated flight path (Y' ) is greater than zero orIf the absolute value of the roll ra e is greater than
80 ° /Sec,
then A Z is set to zero.
rr
48
b. 1 Problems
b.1.1 Early warnings
b.1.1. The algorithm assumes incorrectly that no dive
recovery begin. until roll recovery is complete. In fact, test results
showed that pilots reduced bank and applied recovery "G" simultaneously.
In cases where the aircraft is in a high "G" turn merely rolling out of
the bank while maintaining the "G" affected most recoveries.
b.1.1.2 The use of three distinct roll rates for
predicting roll recovery times causes abrupt changes in the altitude loss
prediction. For instance, the algorithm would predict a T rrof 0.91
seconds for a bank angle of 840. By increasing the bank to 85% T r
jumps to 1.2 seconds. Decreasing the bank from 145* to 144* causes T r
to change from 2.2 seconds to 1.4 seconds. At 450 knots and 30* nose
down the latter time difference (0.8 seconds) is equivalent to an altitude
loss of about 300 feet. At 450 knots and 300 nose down, the latter time
difference (0.8) is equivalent to an attituCde loss of about 300 feet. A
continuous function for calculating T rrfrom 0would eliminate the
abrupt changes and the resultant early warnings.
b.1.2 Late warnings. No late warnings were associated
with the roll recovery c~alculations.
c. Altitude Loss During Dive Recovery
The algorithm calculates the altitude loss that would occur
during a wings level, constant G dive recovery in the following manner:
(1) The basic equation for a constant G constant true airspeed dive
49
recovery is:
~z pu -fV 2jln LIn co Ye~xt]
Where4Z - the altitude lost during the pulluppu
V - aircraft true airspeed (ft/sec)
N - G available (max of 5)
Y'ext = extrapolated flight path angle
g - 32.2 ft/sec 2
This equation is based on the assumption that dive
recovery is complete when Y - 0*
(2) The value for n was determined from a look-up table using the
following piecewise linear schedule (Figure 15).
n (is)
5.0
4.0
2.2
1.8
200 25s 32 425 V (KIAS)
Figure 15 G-Available vs Indicated Airspeed
(3) Since the constant G dive recovery equation does not account for
altitude lost during application of G, it was modified, resulting in the
50
following equation:
AZ p- fK 1 V ln ;n 2 - Csext1
where N 2 anfor n <2.2
a n/2 for n > 2.2
K a IV 1 Cos x + a2 Cos Y'f+a 3 V1 +a 4
V -true airspeed
Vaindicated airspeed
The constants a.i were derived from data obtained by Cubic during a
limited series of controlled dive recoveries in instrumented aircraft.
The equation for A Z puwas assumed to be of the above form and a curve
fit was performed to determine values for constants a.for five
distinct airspeed regimes.
c.1 Problems
c.1.1 Early warnings
c.1.1.1 The equation forA Z pu , along with its variables
a.i and K, was derived based on the assumption that the pilot will start
his dive recovery from a load factor of 1 G. Even during a 5-G dive
recovery, the algorithm continues to predict altitude loss based on this
equation which (as explained in paragraph c (3) above) has been tailored
to include altitude loss during the buildup from I to 5 G's. This can
cause early warnings under certain conditions. Other times, the warning
will stay on after the aircraft is no longer in danger. The designers of
the algorithm have compensated for this inflexibility somewhat by
reducing the altitude loss due to pilot response time (Paragraph a.)
when the aircraft has an upward acceleration, but this does not adequately
solve the problem under all conditions.
51
c.1.1.2 The dive recovery equation calculates altitude
loss in reducing Y to zero. In cases where the terrain is rising, this
results in an excessive estimate.
HI - -------
H2
Figure 16. Error Caused by Estimating AZBased on Recovery to Y= 0 P
Figure 16 shows an aircraft in a dive recovery in the presence of
sloping terrain. The aircraft is nearest the terrain at the point where
the flight path anglerY is equal to the angular slope of the terrain.
This altitude Hl, is higher than the altitude where Y sO, here
represented by H 2. By calculating altitude lost based on reducing Y to
zero, an error of magnitude (H 1- H 2) is introduced, resulting in an
early warning.
This is an important consideration in limiting nuisance warnings due
to rising terrain. In level flight, approaching rising terrain, calcula-
tions based on matching to the terrain slope should result in a negative
AZpu
52
c. 1.2 Late warnings
No late warnings were noted. However, recommend that the
values for K and ai, if used, be verified through more extensive
experimentation.
d. Altitude lose due to terrain rise
The algorithm calculates the altitude loss due to ground rise
using a filtered estimation of the terrain slope directly below the air
craft. To estimate the change in terrain height during recovery, the
slope is extrapolated for a time T which is calculated as a functiongr
of initial flight path angle. The formula for T is:
grtan (n2 + tan 2
L gn2~ ta L+2l1 na
where:
T = time used to calculate ground risegr
V a aircraft velocity (ft/sec)
g - 32.2 ft/sec
n - 4 (G's)
S- aircraft flight path angle
Once T has been determined, the following checks are made togr
determine whether or not to calculate ground rise.
1. Calculate T.impact
T. = _eigh
impact V (terrain) -V (aircraft)z z
where T. - the time it will take the aircraft to
impact hit the terrain based on its current vertical
velocity (Vz(aircraft))and the current rate of
ground rise (V z(terrain))
Height = current height above terrain
53
If Timpact < Treap' ground rise is not calculated.
( T rep - pilot response time, 2 seconds.)
2. Calculate lower boundary for T.impact
BOUNDlwr a Tgr - Timpact
3. Calculate upper boundary for T impact
BOUND m T + 10 Secondsupr gr
4. If Timpact falls between Boundlwr and Bound upr, ground rise
is calculated:
Ground rise (A Z ) is simply calculated as:gr
AZgr (Vz(terrain))- (Tgr
d.l. Other problems
Assuming that the calculated terrain slope is accurate and the
terrain is absolutely consistent, the most accurate way to calculate ground
rise is to multiply the slope of the terrain by the horizontal distance the
airplane will travel during reaction time, roll recovery, and dive recovery
(to match Y to terrain slope). The algorithm calculates terrain rise based
exclusively on the time it takes to perform a constant 4-G dive recovery to
zero flight path angle. This involves several inaccuracies:
(1) Vz(terrain)(ft/sec) must be calculated based on the instantaneous
horizontal velocity of the aircraft dx/dt and the slope dh/dx of
the local terrain, that is:
Vz (terrain) a dh d__dx at-
Since dx increases during the dive recovery, as the aircraftdt
velocity vector nears horizontal, V z(terrain)should increase. Basing the
terrain rise on this initial value of the aircraft's horizontal velocity
54
will result in an underestimation of terrain rise.
(2) The distance traveled during reaction time, roll recovery, and
application of recovery G is not considered. The net effect of these
omissions is an additional nerative error (underestimate) of the terrain
rise.
(3) The calculation of T is based on reducingYto zero. Thus, ifgr
the airplane is in level flight, approaching rising terrain T andAZgr gr
are zero, affording absolutely no protection against rising terrain. If
the "dive recovery" portion of the ground rise calculation considered the
distance traveled while matching Y to the slope angle of the terrain,
more protection would be provided.
(4) From a cursory mathematical analysis of the equation for Tgr
it appears that there may be a sign error in the ternsl n 2 + 1.
This equation appears to be based on the following relation for
constant - g dive recoveries:
dr - g . (n - cosY )dt v
dt -vdYg(n-cos Y"
Integrating both sides of this equation gives:
~Tgr /Y
T = j- ) dt -. Vd(0 0 g(n-cos y)
- 2V tan- tan2
(Ref: CRC Standard Mathematical Tables, Fifteenth Edition, Integral No.260)
Note the term n2' I in place of 1n2.+
This introduces an error as does the use of 4G's instead of 5G's (or G
55
available) for the quantity n.
(5) Since the algorithm takes the absolute value of thequantity 1
tan- [J + 1 tan
n-i
a positive value will be returned for T whether the aircraft isgr
climbing or diving. Thus, an airplane in a slight climb, well
clear of rising terrain will receive a nuisance warnir.g, where
an aircraft in level flight, in danger of striking that same
terrain will receive no warning at all (Figure 17).
(a) (b)This aircraft receives a This aircraft receives nonuisance warning. warning at all.
Figure 17
(6) Since the algorithm uses a linear function to project the change
in terrain elevation, early warnings occur whenever the first
deviative of the terrain slope is negative (Figure 18).
Basically, the linear predictor is too steep for upslopes and
too shallow for downs lopes.
56
- - ~- --- ---- -- -- - ~
4,,
(a) (b)Too sb'allow for downslope Too steep for upslope
Figure 18
Because of previously discussed characteristics of the GCAS
algorithm, very few warnings occurred as the aircraft approached rising
terrain. Howiever, when a "bunting" maneuver was performed (a slight
negative-G pushover) to descend on the back side of a ridgeline, numerous
false warnings were experienced. It is possible that some of the false
warnings could be eliminated if a modified second order curve were used
to predict terrain elevation.
57
LA~WARNINGS
RISING TERIN".LONG STEEP TURN
"* CARA BEYOND LIMITS
"* AIRCRAFT DOES NOT "SEE"RISING TERRAIN
Figure 19 Late Warnings -- Terrain ExtrapoLation
INVALID WARNINGS OCCUR IN LONG,DESCENDING TURNS
LAST CARAUPDAE
- - - ESTIMATED TERRAIN
Figure 20 Nuisance Warnings -- Terrain ExtrapoLation
58
4. TERRAIN ESTIM4AT ION
The GCAS algorithm is designed for use in conjunction with the
Combined Altitude RADAR Altimeter (CARA), which is specified to provide
accuzate altitude whenever the aircraft bank and pitch angles are less
than +60' and ±45 respectively. No forward looking sensor is avail-
able to measure terrain height along the predicted flight path. A
Kalmani Filter provides an estimate of terrain elevation and terrain slope
beneath the aircraft. When the CARA inputs are unavailable, the algorithm
retains the last estimated terrain altitude and extrapolates it according
to its last estimated slope for a period of time, T exts which varies
according to aircraft flight parameters and terrain consistency. When this
time period has expired, the extrapolation is discontinued and the last
terrain height estimate is held until CARA inputs are restored.
This approach allows the GCAS to continue to provide protection over
amuch larger range of flight maneuvers than most alternative systems,
which merely shut down when a current sensor input is unavailable.
However, some problems were encountered during the simulation which
demonstrated limitations to the effectiveness of the terrain extra-
polation/hold feature.
The basis for all the limitations is the fact that terrain is, to a
large degree, unpredictable. If , as in Figure 19, for instance, the air
craft is in a long, high-banked (>60*) turn, and encounters abruptly
rising terrain, no warning will be provided. If the aircraft is in a
descending turn over an abrupt drop off, such as a ridge (Figure 20), an
invalid warning will occur as the aircraft approaches the estimated
terrain altitude. These faults are more or less unavoidable as long as a
59
bank limited sensor with no forward-looking capability is used. Overall,
it appears that in spite of the shortcominf%, the terrain hold/extrapola-
tion feature is a worthwhile addition to the GCAS. In comparision to
systema with no extrapolation feature, GCAS:
1. Provides increased protection during large amplitude maneuvering
over flat or regularly sloping terrain.
2. Increases the number of nuisance warnings slightly due to turning
descents over irregular terrain.
3. Provides the same proteccion against steep turns into rising
terrain--none.
60
I
5. RECOM[ENDATIONS
There is no question that Cubic's approach to the GCAS process is
most ambitious and promising. The concept of terrain estimation, both for
periods when the sensor inputs are degraded and for prediction of altered
loss due to upslope, may give GCAS a capability unmatched by any system
using similarly limited sensors.
Recommend, however, that the preceding analysis be considered carefully
priir to implementation of the system. Follow-on testing in actual air-
craft should be used to evaluate the Kalman Filter and its capability with
a CARA system over actual terrain. These portions of the system could not
be properly evaluated in the si-mulator due to (1) dissimilarities between
simulator terrain data base and actual terrain and (2) uncertainties about
the true characteristics of CARA. If the limitations mentioned herein are
adequately remedied, then the generic GCAS stands a very good chance of
saving valuable aircraft and irreplaceable crews.
61
6. RESUJLTS
In order to integrate the pilot performiance/response time data with
the GCAS algorithm Analysis and pilot questionnaire, a brief summary is
appropriate. First, the duration of the GCAS warning varied depending on
type of mission flown. The GCAS duration ranged from 1.85 seconds for
low level navigation to 4.38 seconds for I.HC unusual attitudes. The
average warning duration across all missions was 3.1 seconds
(S.D. - 2.34). As is the case with most response time daca, it tends to
be slightly skewed due to a variety of factors that were discussed.
The average response time to a GCAS warning is 0.66 seconds with a
standard deviation of 0.30. Taking all factors discussed into account
and adding 2 sigma to the mean response time, the recommnended allowance
for pilot response in the GCAS is 2 seconds. Performance data were also
gathered in time to maximaum G (M =2.68; S.D. - 1.39), maximuim G's pulled
CM - 4.62; S.D. - 2.34), and time to zero flight path angle (M - 3.88;
S.D. - 2.56).
The Algorithm Analysis indicates some modifications are necessary
including the following:
1) Modifications for smoothing roll, roll rate and acceleration.
2) Shortening flight path extrapolation for negative G "bunts."
3) Adjusting roll recovery rates and varying them with airspeed and
bank.
4) Modifying method of computing terrain rise.
5) Modifying reaction time allowance at zero flight path angle.
Finally, the questionnaire data indicated a good pilot acceptance and
the results indicate that GCAS is a viable system for partial solution of
CFIT.
62
SECTION IV
CONCLUSIONS
1. A generic GCAS appears to be a useful system that has good pilot
acceptance, but needs some improvements prior to implementation.
2. The recommended improvement to the algorithm should be
incorporated and verified in simualation (see recommendations).
3. Flight test the improved system.
63
GROUND COLLISION AVOIDANCE SYSTEM PILOT QUESTIONNAIRE
INSTRUCTIONS: The purpose of this questionnaire is to obtain information
from you about your previous flying experience. Your answers to thesequestions will help us in our evaluation of this simulation. Your honest
opinions are, therefore, essential and will be kept confidential.
If you have any questions, please ask the questionnaire administrator for
assistance. Take as much time as necessary to answer the questionnaire.
PERSONAL DATA:
Name (last, first, mi): ______________ ___________
Rank:
Duty AFSC: ____________________________
Organization and Symbol: _____________________
Duty Station:____________________________
Duty Phone:___________________________
Wing Commander, Squadron Commander, DO: ______________
Aero Rating: ____________________________
Age:-
Height:________________________________
Weight: __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Vision, Corrected: ___________Uncorrected:_________
Years in Military Service:_______________________
What type(s) of aircraft have you flown? (List)
Aircraft: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Fighter hours: __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
F-16 Hours: _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Hours Flying with Radar Altimeter:________________
Total Flight Time (Include Student Hours):____________
64
P ILOT QUESTIONNAIRE
1. Do you feel the "pull-up" warning alarms were initiated for the most part:
1. Too early2. Slightly early3. Just Right4. Slightly late5. Too late
2. Did the warning system get your attention?
1. Yes, all the time2. Yes, most of the time3. Occasionally4. Seldom5. Never
3. How often did you get nuisance warnings?
1. All the time2. Most of the time3. Occasionally4. Seldom5. Never
4. Were there any instances, in your opini..n, where the system providedno warning when it should have? Describe the circumstances.
5. What do you feel is the usefulness of a GCAS system?
I. Extremely useful2. Useful3. Moderate usefulness4. Useless5. Extremely useless
65
6. How often, if ever, would you tend to over rely on a GCAS warningsystem (i.e., using the system as a pull-up cue during a botdbingmission)?
1. All the time2. Frequently3. Occasionally4. Seldom5. Never
7. How would you rate the adequacy and suitability of the GCAS warning?
1. Excellent, optilmum warning system2. Very good3. Good4. Satisfactory, but improvements could be made5. Satisfactory, but improvements are essential6. Poor7. Very Poor
8. Were there any design considerations omitted that you consider essentialto GCAS ?
9. General comments on GCAS_______________________
P 10. Comment on the adequacy of this simuilation for rating a GroundCollision avoidance system algorithm. _______________
66
QUESTIONNAIRE DATA
Q. (1.) Were there any instances, in your opinion, where the systemprovided no warning when it should have? Describe thecircumstances.
Pilot No. 1: Yes, on the low level mission (No. 11), 1 skimmned the ground
with no warning at the very beginning.
Pilot No. 2: No.
Pilot No. 3: No.
Pilot No. 4: No.
Pilot No. 5: No.
Pilot No. 6: No.
Pilot No. 7: No.
Pilot No. 8: No.
Pilot No. 9: When flying into rising terrain (but it's not meant toprovide that protection). It also seemed slow to reactto a level bunt into flat terrain.
Q. (.. Were there any design considerations omitted that you
consider essential to GCAS?
Pilot No. 1: No.
Pilot No. 2: Only one, maybe looking out in front of the jet some wouldhelp, if possible.
Pilot No. 3: None.
Pilot No. 4: One improvement discussed would provide an optical vernierrepresentation of an approaching pull-up call(i.e., how close am I to getting a pull-up call).
Pilot No. 5: Pilot warning as to need to pull-up prior to it beingcritical to pull.
Pilot No. 6: More timely voice.
Pilot No. 7: Wider looking radar Altimeter. Predictive curvature atranges f or ridge crossings (i.e., when inverted ridgecrossing needs to predict curve).
Pilot No. 8: Yes, there should be a method, such as foward radaraltimeter, to include terrain in front in the algorithm.
67
This would eliminate most, if not all, nuisance warnings.
Pilot No. 9: A clear signal distraction when GCAS comes on in the AUTO
mode. Brief glitches leave you wondering what happened.
Q. (3.) General comments on GCAS.
Pilot No. 1: System works well as long as no trim inputs caused. Need tocorrect ridge crossings so pilot doesn't get too manynuisance warnings during ridge crossings.
Pilot No. 2: Seemed to work qA.te well. Only problem was with nuisancewarnings after ridge crossings.
Pilot No. 3: A good system, needs further testing/changes for falsewarnings after ridge crossings. Also, false warnings whenflying alongside a ridge. It appeared that the radaraltimeter was looking sideways.
Pilot No. 4: Good system. Due to less than optilmum VIS, the systemappeared to have good potential. If the pilot can turn thesystem on/off, it would be great in some circumstances.That is, missions flown into the sun in the desert wherehills blend into the shadows, etc.. Ridge crossings seemto trip the system too much.
Pilot No. 5: Manual system worked pretty well as a heads up call to adistracted pilot; even though the system had problems onthe downhill side of a r- dge crossing (false warning).
Pilot No. 6: Needs work on ridge crossings.
Pilot No. 7: Good concept. Eliminate false warnings and will be helpful.
Pilot No. 8: GCAS is essential. I think this system is a good start.Ithink you could put this present algorithm in the F-16 todayand it would be adaquate, although pilots would be apt tocomplain about the nuisance calls. I theorize pilots mightactually fly their aircraft differently so as to minimizenuisance warnings. The only way to eliminate these falsewarnings is to have a foward-looking source as well as thestandard radar altimeter. One other comment: the twosecond reaction time may be too long in the algorithm. Itwas hard to uitmulate this since I was expecting many of thewarnings but in the aircraft, T* think a pilot might reactslightly quicker than two seconds to an unexpected warning(pull-up now, ask questions later). Over all, it's a goodsystem!
Pilot No. 9: Needs to be given without delay. The false warnings onridges were less frequent than I expected so the warningsthat did come on, occasionally took me by surprise.
68
Q. (4.) Commnt on the adequacy of this sitmulation for rating aGround Collision Avoidance System algorithm.
Pilot No. 1: Will cut down on CFIT, but I don't think the system willcompletely eliminate the ground collision problem.
Pilot No. 2: Very good. Naturally, since you can't have perfect visualsitmulation you can't duplicate conditions perfectly, but itseemed to work fairly well.
Pilot No. 3: None.
Pilot No. 4: Good. The lack of periferal vision makes turning at lowaltitude and performing additional tasks tough.
Pilot No. 5: Very good. The simulator flies much like the jet.
Pilot No. 6: Good.
Pilot No. 7: Good.
Pilot No. 8: Fairly poor. Biggest problems: visual perceptions ofdepth, lateral position, and poor lighting. It is quitepossible that one may have almost no nuisance warnings ifthe aircraft was flown (over).
Pilot No. 9: The profile was great.
Q. (5.) How does the "slope" warning compare to the pull-up call?
Pilot No. 1: Good for level terrain, but useless with really steepridges.
Pilot No. 2: 1 like it well enough, but at low altitude it could give youa warm feeling falsely when you in fact should be getting a"Pull-up." Could be very useful for insidiously risingterrain.
Pilot No. 3: Pull-up is all that is necessary.
Pilot No. 4: 1 like "UPSLOPE." It is less harsh, a remark rather than acommand.
Pilot No. 5: The slope call seemed kind of useless in that it came onabout one second before the pull-up command. The upslopeshould either occur imuch earlier or not at all. (That is tosay, only have the pull-up command.)
Pilot No. 6: Use pull up all the time.
Pilot No. 7: No value. Either call brings your attention inside andevaluate the situation(i.e., you are not going to start andpull unless you know your~ attitude).
Pilot No. 8: Not necessary. It defines too narrow a range of conditions,
69
and almost overlaps the already adequate pull-up call.Besides, pilot reaction would be the sam, so why use twowarnings?
Pi lot No. 9: 1 did not see imuch utility in the elope. The actions werethe same, so I would stay with one.
70
-~~~~~~~- Pon - -.- ~-- ~ ~
TEST PROCEDUREMON TUE WED, THU FRI
WEEK 1 2 TEST 2 TESTA.M. MISSIONS MISSIONS
P.M . IN-BRIEF 2 TEST 2 TESTAND TRAIN MISSIONS MISSIONS(PILOTS I ANDAND 2 ) IDEBRIEF
WEEK 2 IN-BRIEF 2 TEST 2 TEST 2 TEST 2 TESTA.M. AND TRAIN MISSIONS MISSIONS MISSIONS. MISSIONS
(PILOTS 3 ANDAND 4 )DEBRIEF
P.M. 2 TEST 2 TEST IN-BRIEF 2 TEST 2 TESTMISSIONS MISSIONS AND TRAIN MISSIONS MISSIONS
(PILOTS 5 ANDAND 6 )DEBRIEF
WEEK 3 IN-BRIEF 2 TEST 2 TEST 2 TEST 2 TESTA.M. AND TRAIN MISSIONS MISSIONS MISSIONS MISSIONS
(PILOTS 7 ANDAND 8 )DEBRIEF
P.M. 2 TEST 2 TEST IN-BRIEF 2 TEST 2 TESTMISSIONS MISSIONS AND TRAIN MISSIONS MISSIONS
(PILOTS 9AND 10)
71
MISSION BRIEFINGS
MISSION 11 - LOW LEVEL NAVIGATION
OBJECTIVE: Evaluate GCAS algorithm for nuisance warnings/pull-ups duringaggressive low-level maneuvering.
SCENARIO: This is a low-level nav mission planned at 300 to 500 feet,480 knots. Fly the nay profile as closely to theplanned line as possible. Threats are active above800 AGL. Fly as aggressively as you can to stay below
500 feet. You have programed TOS's for each steerpoint.
START: Altitude: 2200 MSLGroundspeed: 480 KnotsHeading: 210
INSTRUCTIONS:
1. Make your TOS (Groundspeed cue will help).2. Stay below 800' or get zapped!3. Evaluate each "pull-up" call as:
A. ValidB. Invalid, expectedC. Invalid, unexpected
72
MISSION 12 - LOW SPEED MANEUVERING, INC
OBJECTIVE: Evaluate GCAS low speed effectiveness against varied terrain.
SCENARIO: You have just completed an enroute descent to BuzzardCreek AFB. Since you lost the approach plate, you haveasked for radar vectors to a PAR final. The controller hasassigned you an altitude of 1800 feet MSL. That sounds prettylow, but you think you remember that the MSA was 1400 feet.You are in IMC. Your radar altimeter is busted.
START: Altitude: 1800 feet MSLAirspeed: 250 KCASHeading: 268 degrees
INSTRUCTIONS:
1. Maintein 250 KCAS and follow the steerpoint cues.2. If you get a pull-up alert, pull as hard as you can--use
power as you feel the situation dictates.3. Climb to 3000 MSL for 15 seconds, then descend expeditiously
to 1800 feet again.
73
MISSIOIN 13 - HARD TURMS AT LOW ALTITUDE
OBJECTIVI. Evaluate GCAS effectiveness during hard defensive maneuveringat low altitude.
SCENARIO: You are practicing hard defensive turns at low altitude.You will fly a figure 8 pattern between Millersburg andElizabethville. To keep you honest, a "SAM" threat will soundeach time you go above 800 AGL.
START: 1000 feet MSL (500 AGL)Heading: 094Airspeed: 480
INSTRUCT IONS:
1. At each steerpoint, make a hard (5-G) turn back towardthe next steerpoint.
2. Always turn toward the north, away from the mountains.3. Each time you turn, you will be instructed to read a
series of numbers and letters from a card located aboveyour head to simulate checking 6. Try to maintain yourG-load and altitude while you read the data.
4. If you get a warning, recover and continue.
74
MISSION 14 - RANGE MISSION
OBJECTIVR. Evaluate GCAS effectiveness during a complex range missionwith side taskittg.
JCEPURIO: This is your first trip to the Berrysburg Bombing Range.You are trying out a number of alternate run-ins forstraight ahead pop-up deliveries. 3ecause of a peculiarlocal airspace restriction, you must remain at or below500 AGL except in the target area.
START: Altitude: 1300 feet KSLAirspeed: 490 knotsHeadicig: 199
Weaporxs: 4 MK 82's. stns 3 7,CCIP/Singles RP - 1
INSTRUCTIONS:
1. Make 4 straight ahead pop-up attacks on the Berrysburgtarget.
2. The sequence of the runs is on your map. After each run,fly to the bend in the river and then along Hoof landerMountain.
3. If yziu are in doubt, follow the steerpoints. Consoleoperator will help.
4. Make tight turns (at least 3 G's).5. Start your pull-up when you are 20 seconds from target.
- pull up 30 degrees.- pull down at 4000 feet MSL (30 degrees).- release at 2500 feet MSL.- do your best uot t. sbort the pass -- even if you
feel a bit steep (unless you get a pull-up warning).6. Stay in the NAY Master Mode until you are on run-in heading,
then select A/G.7. You will be asked to perform additional tasks during the
mission.8. If you get a pull-up warning, recover and -ontinue the
mission.
75
MISSION 15 -LOW LEVEL NAV/POP-UP DELIVERIES
OBJECTIVE: Evaluate the GCAS effectiveness during low altitude tacticalnavigation/veapons delivery.
SCENARIO: This is a combined nav/weapons delivery profile. There arethree targets planned for straight ahead pop-up deliveries.The targets are fairly easy to see: a bridge, a ship, andan airfield. The area is heavily defended. If you climbabove 800 feet AGL, you will get a "SAM" warning.
START: Altitude: 2100 feet NSL (500 feet AGL)Airspeed: 480 knots
Heading: 054Weapons: 6 Hk 82's/Pairs/RP I/CCIP
INSTRUCTIONS:
1. Fly 480 knots G's or the speed required to make yourTOT. Turns are 60 degrees bank. Stay at 300 AGL
except during weapons delivery. Maintain 480 feetduring target run on.
2. Weapons Delivery:
TGT Pull-Up Pull-Down ReleaseI (Bridge) 30 degrees at 20 sec 4100 feet 2600 feet2 (Ship) 30 degrees at 20 sec 4000 feet 2500 feet3 (Airport) 30 degrees at 20 sec 4000 feet 2500 feet
3. If you get a pull-up warning, recover and continue themission at the planned altitude.
4. If you get a warning during NAV leg, evaluate it aseither:
A. Valid5. Invalid, expectedC. Invalid, unexpected
76
MISSION 16 - INC MANEUVERING
OBJECTIVE: Evaluate the system's effectiveness in recovering fromsituations where spatial disorientation is a factor.
SCENARIO: You will perform a number of aerobatic maneuvers in INC.
NOTE: A non-standard HUD Bank display will be used to help yourattitude awareness.
INSTRUCTIONS:
I. Stay within 10 nui of steerpoint "1". Try to work back andforth between STPT I and 2.
2. Perform the following maneuvers:
A/S ALT (feet) G's POWER
Split S 300 8000 G's and power as required tomaintain 300 knots.
Loop 500 3000 5 NIL
Cuban 8 500 3000 5 NIL
Sliceback 350 8000 3-5 NIL
Double 300 14000 G's and power as required toSplit S maintain 300 knots.
3. If you get a pull-up warning, complete the recovery toa safe altitude above 500G feet and set up for the nextmaneuver.
77
REFERENCES
I. Linquist, ER.., Design and Analysis of Experiments in Psycholosy and in Education.Boston: Roughton Mifflin Co., 1953.
2. ASD/AG, Draft Statement of Work, 14 March 1986.
3. Cubic Corporation, Feasibility Demonstration of a Gra ind Collision Avoidance System,Technical Report 414-1A, 1 September 1985.
78